Trends in total and cause-specific mortality by marital status among elderly Norwegian men and women

Figures 1, 2 and 3 show relative differences in mortality by marital status from the selected causes of death for the years 1971-1979, 1980-1989, 1990-1999 and 2000-2007. In all models there are controls for year, age, educational level, region of residence and centrality. There are pronounced differences in all-cause mortality by marital status, particularly for men (Figure 1). The mortality gap is especially large between divorced and married persons. From 1971 to 2007 the excess mortality compared to the married has increased for most groups of non-married, and most clearly for the never married. Moreover, the increase has been more pronounced for women than for men. In the time period 2000-2007, divorced, never married and widowed women had almost as high excess mortality relative to those who were married as their male counterparts.

For all cancers together, marital status differentials have been fairly stable over time (Figure 2). However, never married men have significantly lower risk of dying relative to their married counterparts in the beginning of the study period, and an increasing advantage from 1971-79 to 1980-1989. The pattern varies across cancer types, though (Figure 2). There are few significant marital status differentials in mortality from colon/rectum cancer and no obvious time trends. Apparently there is a marked drop in the excess mortality for divorced men over time. However, their odds of death are not significantly different from that of married men from the 1980s. With respect to lung cancer mortality, never married men are at a significantly lower level than those who are married, whereas divorcees have a significant excess mortality. The gap is more or less stable over time. Differences in lung cancer mortality between divorced and married women have been reduced over time, while there has been a slight increase in the excess mortality of widows. Significant differences in mortality from prostate cancer are only found between never married and married men in the beginning of the study period, where the first group shows a lower risk, while divorcees and widowers have significantly higher mortality levels as compared to married men only in the third time period. The risk of dying from breast cancer is significantly higher among never married women relative to those who are married, and the differences have increased since the 1980s.

There are large and increasing marital status differentials in mortality from circulatory diseases (Figure 3). The increase is most pronounced for ischemic heart diseases, whereas the increase in the marital status differentials in cerebrovascular mortality has been more modest (results not shown).

In general, marital status differences in mortality from respiratory diseases are large (Figure 3). However, time trends are different for males and females. For males, there has been a convergence over time due to lower excess mortality among divorced men, whereas females have experienced increased relative differences in all non-married groups. Large and increasing differences in mortality by marital status are also found for mortality from other diseases and external deaths (Figure 3). The exception is divorced men, for whom there is no clear trend in the excess mortality with respect to external deaths.

Table 3 shows linear trends in the association between marital status and all-cause mortality (i.e. interactions between marital status and period) for the whole time period 1971-2007. There is a significant increase in excess all-cause mortality for all non-married groups relative to those who are married, with no substantial gender differences, except for the divorced group where the increase is statistically significant only among women. The highest excess mortality is among the never married group, with an increase of 0.8-0.9% per year from 1971 to 2007, respectively for men and women. Significantly increasing mortality differentials are also found for deaths from cancers (an increasing excess mortality for never married males and females and widowed males), circulatory diseases, respiratory diseases (an increasing excess mortality for widowed males and all females), other diseases (except for divorced men) and external causes (an increasing excess mortality for widowed males and females and never married females) (results not shown).

The results are similar to those reported in several other studies, which have also concluded that elderly non-married persons have an excess mortality from all-cause mortality [1‐6, 13, 14], cancers [2, 14], cardiovascular mortality [2, 6, 14], respiratory diseases [6] and external deaths [6].

The higher mortality found for most causes of death among elderly non-married Norwegians may be due to protective effects of marriage. First, married persons are likely to benefit from various types of support [15]. A spouse may exert control on behavior, offer practical help, add to the pool of knowledge and help interpreting important information. Second, having a spouse is an economic advantage because of specialization, economies of scale and pooling of wealth [16]. Through either channel, marital status may be associated with various lifestyle factors of importance for the occurrence of a number of potential lethal diseases. Examples of such lifestyle factors are smoking, which increases the risk of lung cancer, cardiovascular and respiratory diseases [17], overweight and obesity, which are associated with a high risk of cardiovascular diseases and various cancer types, including breast cancer, colorectal cancer and aggressive prostate cancer [18, 19], physical activity which lowers the risks for several diseases, and alcohol consumption, which increases the risk of external deaths, liver diseases and possibly breast cancer [20]. Moreover, some diseases are associated with reproductive factors that are linked with marital status. For example, low sexual activity is associated with low occurrence of prostate cancer [21], and this may be one reason for the lower mortality for this cause of death observed among never married men in the beginning of the study period. Moreover, the high excess mortality from breast cancer found among never married women may be partly due to childlessness as childbearing has a protective effect through physiological mechanisms [20].

Protective effects of marriage may also operate through factors affecting survival from different diseases. First, those who are married may get diagnosed and seek treatment earlier, and therefore have a better prognosis [22]. This is especially important in case of cancer, where the results show significant marital status differences in mortality. Second, health status at time of diagnosis and health behavior in the subsequent period are important predictors of survival and related to marital status through the mechanisms mentioned above. Third, those who are married may get better treatment e.g. because their spouse help them to ask for a second opinion, discuss possible treatments and help them follow the given instructions [23].

In addition to being associated with current marital status, mortality may be related to the time since the most recent change in marital status as well as the earlier history. In particular, some of the previously married may have experienced death of the partner quite recently, followed by stress reactions that add to the aforementioned problems associated with being non-married. This may be most relevant for external causes of death, where the results show high excess mortality, especially for men. In support of this kind of explanation, a Finnish study has shown that excess mortality is especially high for accidental and violent causes of death shortly after having become widowed, even among elderly people [24].

Another reason for the marital status differentials in mortality is that certain characteristics affect both mortality and family behavior. Most importantly, poor physical or mental health typically reduce the chance of marrying and remaining married, though the opposite is also possible, because presumed protective effects of marriage may give persons in poor health incentives to marry [25]. Furthermore, there may be a selection into or out of marriage with respect to characteristics affecting health behavior, risk-taking behavior or emotional stability, such as socio-economic status or values [26].

A striking result is the lower risk of dying from lung cancer among elderly never married men through the whole time period. However, this is in accordance with findings from a Dutch study based on the entire population of 25 years and older [27]. Lower lung cancer mortality among never married men may be due to smoking behavior as some studies suggest that the proportions of smokers among the never married are relatively low [28, 29]. It is less likely that the lower mortality can be explained by health selection over age, as these patterns also were present in younger age groups (results not shown).

The results show increasing marital status differences in all-cause mortality for elderly Norwegian males and females. This is in line with findings from other studies focusing both on older people and on younger age groups [3‐6, 14, 30]. I was able to identify only a few studies investigating trends in cause-specific mortality by marital status. In accordance with findings from Israel for the age group 65-89 [14] and the US for ages over 40 [30], but as opposed to the pattern reported from Finland [6], this study did not show any increase in marital status differences in cancer mortality. The results for cardiovascular mortality are similar to those reported from Israel [14], the US [30] and Finland [6], and show increasing differences by marital status over time. In accordance with the findings from Martikainen et al. [6], there were divergent trends in female mortality from respiratory diseases, whereas excess mortality for divorced males decreased over time, leading to an overall reduction in the marital status differences for men. Contrary to the results from the US [30] but similar to those from Finland [6], widening marital status differentials in external deaths were observed. Moreover, the increase in marital status differences in all-cause mortality in the current study was most pronounced for women, and the gender differences in mortality by marital status were almost equal in the end of the study period.

There may be different reasons why associations between marital status and mortality have changed over time among elderly Norwegian men and women. First, the increasing excess mortality among never married persons evident for most causes of death may indicate that they receive less support from others as the proportion never married, among whom they perhaps are most likely to have their closest companions, has become smaller. Social support from persons sharing the same situation may be most important for never married persons, as divorcees and widows/widowers are more likely to have children. Moreover, there is probably a stronger focus on self-realization today, and people may care most for their closest family. This may have played a role for the reported increase in excess mortality among the non-married in general. Second, the economic benefits of marriage were mostly due to specialization earlier-with the female partner having responsibility for home and child care while the man was the only wage earner [16]. It has been suggested that the economic benefit now to larger extent lies in the pooling of resources and advantages of scale e.g. [31], and one may speculate whether this perhaps is less of an advantage. The literature is not conclusive on this topic, but some report lower economic gains from marriage over time, at least for men [32]. However, any such change may be of modest relevance for the cohorts in focus of the current study, among whom women's labor force participation has still been low. Third, it may have become more accepted and thus less stressful, to be divorced as this status has become more common [33]. This may suggest why the disadvantage for divorced men has been stable over time, and even reduced for some causes of death. Fourth, there may be changes in selection into and out of marriage over time. Some studies have focused on changes in relative group size and argue that decreasing shares of non-married persons indicate stronger selection with respect to health and cultural or socioeconomic characteristics that have a bearing on health e.g. [34]. This may be the case for never married elderly people. Their relative group size has decreased from 1971-2007, whereas their excess mortality relative to those who are married has increased. However, changing proportions in the non-married and married population is a too vague indication of the changing direction of the selection. The important issue is whether characteristics that have a negative effect on health or health behavior have become more or less important for marriage formation and dissolution. Unfortunately, there is little knowledge about this. For example, we do not know whether it has been considered increasingly important to find a partner who has good health or a high wage potential, or who is a good problem-solver, all of which would be potentially negatively linked to later mortality. In particular, the relative-size argument is irrelevant for the widowed, because a smaller group of widowers is due to increasing life expectancy over time and changing sex differentials in mortality.

Even though several causes of death have common risk factors, investigating trends in cause-specific mortality may indicate how associations between marital status and risk factors for different causes of death have changed over time. For example, increasing excess mortality from circulatory diseases among the non-married may be due to changing diet more than changing smoking patterns, as there is no increase in marital status differentials in lung cancer mortality. According to one Swedish and one Finnish study, healthy diet became more common among married than non-married women over time [35, 36].

There are some limitations in the current study. First, interpreting trends in cause-specific mortality is in principle problematic because of changes in ICD revisions and coding practices over time. However, this is less reason for concern when the focus is on broad causes of death like here. Second, information on cause of death from death certificates is less reliable among elderly people [37]. This is particularly the result of co-morbidity, which is especially common among older persons and perhaps above all those with low levels of social support [38]. In other words, cause of death may be most inaccurate in the non-married population, and can bias the estimates for these groups. Third, the only socioeconomic factor controlled for in this analysis is the educational level, which is relatively homogenous in the age groups in focus. It has been suggested that socioeconomic status among elderly people also should be measured with income and former occupation to give a more comprehensive picture [39]. Fourth, there is no information on cohabitation status as far as back to 1971 in the Norwegian registers. Those living with a partner, friend or relative without being married are registered in one of the non-married groups. When data on cohabitation status is lacking and this group is large, differences in mortality by marital status would be underestimated, as the mortality e.g. among those living with a partner without formal marriage is lower than among other non-married groups [40]. However, there are few persons living with a partner without a formal marriage in the old age groups in focus in the present study. Survey data from Statistics Norway shows that this group constitute less than two percent in the age group 70-79 from 1993-2007 [41].

Despite some limitations, this study contributes to the debate on differentials and trends in mortality by marital status because of the long study period, the large data material and the focus on cause-specific mortality. The reported increase in excess mortality among never married, divorced and widowed persons should be monitored as this points towards a large demand for health care. Moreover, the results from the cause-specific analyses indicate that there is a need to investigate whether marital status differentials in risk factors associated with circulatory diseases, respiratory diseases and external deaths, in which the increase in excess mortality among non-married was most pronounced, have changed over time.